Precision identification and prediction of high mortality phenotypes and disease progression pathways in severe malaria without requiring longitudinal data

Abstract

AbstractThe parasite Plasmodium falciparum is the main cause of severe malaria (SM). Despite treatment with antimalarial drugs, more than 450,000 SM deaths are reported every year, mainly in African children. The diversity of clinical presentations associated with SM indicate important differences in disease pathogenesis that often require specific treatment, and this clinical heterogeneity of SM is largely unresolved. In this study, we apply new machine learning and inference tools for large-scale data analysis to dissect the heterogeneity in patterns of clinical features associated with SM in 2,695 Gambian children admitted to hospital with Plasmodium falciparum malaria. This quantitative analysis, including the powerful HyperTraPS algorithm for inference of progressive processes, reveals pathways of SM symptom progression and features predicting the severity of individual patient outcomes. Notably, our approach allows the identification and dissection of disease progression pathways without the need for longitudinal observations. Learning these pathways and features from this rich dataset allows us to construct several quantitative measures of the mortality risk associated with a patient presenting with a given set of symptoms. By independently surveying expert practitioners, we show that this data-driven approach agrees with and expands the current state of knowledge on malaria progression, while simultaneously providing a data-supported framework for predicting clinical risk.